Donate Bitcoin

Donate Paypal


PeakOil is You

PeakOil is You

Statistics Question

What's on your mind?
General interest discussions, not necessarily related to depletion.

Re: Statistics Question

Unread postby emailking » Thu 08 Jan 2009, 23:41:45

Koyaanisqatsi is right.

P(B|A) doesn't assume A must cause B, even if that's how one usually thinks about this rule. It could cause it, or it might not cause it.

Let C = B happens as caused by A

If B must be caused by A, then P(C) = P(B|A). But if B doesn't have to be caused by A then P(C) < P(B|A). B can be happening irrespective of A.

There aren't fundamentally different rules for independent and dependent events. It's just that if A and B are independent then P(B|A) = P(B) and so P(A and B) reduces to P(A)*P(B)

Note P(A and B) = P(B)*P(A|B) in addition to P(A)*P(B|A). Doesn't matter which causes which, if any.
User avatar
emailking
Tar Sands
Tar Sands
 
Posts: 742
Joined: Sat 11 Mar 2006, 04:00:00

Re: Statistics Question

Unread postby jbeckton » Fri 09 Jan 2009, 20:21:13

$this->bbcode_second_pass_quote('emailking', 'K')oyaanisqatsi is right.

P(B|A) doesn't assume A must cause B, even if that's how one usually thinks about this rule. It could cause it, or it might not cause it.


Which point are you refering to? I don't recall anyone arguing against this.
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00

Re: Statistics Question

Unread postby emailking » Fri 09 Jan 2009, 21:26:41

$this->bbcode_second_pass_quote('jbeckton', '')$this->bbcode_second_pass_quote('emailking', 'K')oyaanisqatsi is right.

P(B|A) doesn't assume A must cause B, even if that's how one usually thinks about this rule. It could cause it, or it might not cause it.


Which point are you refering to? I don't recall anyone arguing against this.


Well Carlhole did: "Falling down stairs is an event. Breaking a leg in a fall down the stairs is a conditional event. You can't use the chances of breaking a leg for ANY reason in this calculation because breaking a leg is conditional on falling down stairs in this example. "

But in any case, it seems to be the way everybody is understanding the concept P(A and B) = P(A)P(B|A) though Koyaanisqatsi isn't using it necessarily in a conditional sense.
User avatar
emailking
Tar Sands
Tar Sands
 
Posts: 742
Joined: Sat 11 Mar 2006, 04:00:00

Re: Statistics Question

Unread postby Koyaanisqatsi » Sat 10 Jan 2009, 04:12:51

$this->bbcode_second_pass_quote('jbeckton', '')$this->bbcode_second_pass_quote('Koyaanisqatsi', '
')And you posted one post here and in the second post you bailed.


Yes, I was interested in having a discussion to clarify my (and your) point, not to listen to you whine about who was right and who was wrong.

There is a difference.


It's not about right and wrong. It's about the Truth - the person who understands the Truth is right and the person who doesn't is wrong. Clearly you are not interested in the Truth and so you are wrong.
User avatar
Koyaanisqatsi
Lignite
Lignite
 
Posts: 351
Joined: Mon 17 Sep 2007, 03:00:00
Location: Pacific NW
Top

Re: Statistics Question

Unread postby Koyaanisqatsi » Sat 10 Jan 2009, 04:27:22

$this->bbcode_second_pass_quote('jbeckton', '')$this->bbcode_second_pass_quote('emailking', 'K')oyaanisqatsi is right.

P(B|A) doesn't assume A must cause B, even if that's how one usually thinks about this rule. It could cause it, or it might not cause it.


Which point are you refering to? I don't recall anyone arguing against this.


This:
$this->bbcode_second_pass_quote('', 'T')here aren't fundamentally different rules for independent and dependent events. It's just that if A and B are independent then P(B|A) = P(B) and so P(A and B) reduces to P(A)*P(B)


which is to say that the argument is valid irrespective of the equation used which is what I have been telling you all along.
User avatar
Koyaanisqatsi
Lignite
Lignite
 
Posts: 351
Joined: Mon 17 Sep 2007, 03:00:00
Location: Pacific NW
Top

Re: Statistics Question

Unread postby Koyaanisqatsi » Sat 10 Jan 2009, 04:27:59

$this->bbcode_second_pass_quote('emailking', 'K')oyaanisqatsi is right.

P(B|A) doesn't assume A must cause B, even if that's how one usually thinks about this rule. It could cause it, or it might not cause it.

Let C = B happens as caused by A

If B must be caused by A, then P(C) = P(B|A). But if B doesn't have to be caused by A then P(C) < P(B|A). B can be happening irrespective of A.

There aren't fundamentally different rules for independent and dependent events. It's just that if A and B are independent then P(B|A) = P(B) and so P(A and B) reduces to P(A)*P(B)

Note P(A and B) = P(B)*P(A|B) in addition to P(A)*P(B|A). Doesn't matter which causes which, if any.


Thanks, very clearly put.
User avatar
Koyaanisqatsi
Lignite
Lignite
 
Posts: 351
Joined: Mon 17 Sep 2007, 03:00:00
Location: Pacific NW
Top

Re: Statistics Question

Unread postby Koyaanisqatsi » Sat 10 Jan 2009, 04:35:38

$this->bbcode_second_pass_quote('Koyaanisqatsi', 'A')s I said, even if you say you meant both=P(B|A), then the statement doesn't map onto my argument about the 9/11 probabilities, so your claim that this invalidates the argument is itself invalid. I would think that someone who genuinely wants to discover the truth would follow an argument to its logical conclusion, but you clearly aren't interested in the truth.


And as usual JBeckton ignores the substance of the post. Dodge and weave!
User avatar
Koyaanisqatsi
Lignite
Lignite
 
Posts: 351
Joined: Mon 17 Sep 2007, 03:00:00
Location: Pacific NW
Top

Re: Statistics Question

Unread postby jbeckton » Sat 10 Jan 2009, 19:38:45

$this->bbcode_second_pass_quote('Koyaanisqatsi', '
')$this->bbcode_second_pass_quote('', 'T')here aren't fundamentally different rules for independent and dependent events. It's just that if A and B are independent then P(B|A) = P(B) and so P(A and B) reduces to P(A)*P(B)


which is to say that the argument is valid irrespective of the equation used which is what I have been telling you all along.


Did you misinterpret another post?

$this->bbcode_second_pass_quote('', 'I')t's just that if A and B are independent then P(B|A) = P(B) and so P(A and B) reduces to P(A)*P(B)


Therefore if the events are NOT independent (such as breaking you leg AND falling down the stairs) then you can NOT reduce P(A and B) to P(A)*P(B).
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00
Top

Re: Statistics Question

Unread postby jbeckton » Sat 10 Jan 2009, 19:43:41

$this->bbcode_second_pass_quote('Koyaanisqatsi', '
')And as usual JBeckton ignores the substance of the post. Dodge and weave!


:lol:

As usual, you presume there is substance in your straw grasping.
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00
Top

Re: Statistics Question

Unread postby emailking » Sat 10 Jan 2009, 21:16:17

$this->bbcode_second_pass_quote('jbeckton', '
')Therefore if the events are NOT independent (such as breaking you leg AND falling down the stairs) then you can NOT reduce P(A and B) to P(A)*P(B).


That's not needed to prove his point though.
User avatar
emailking
Tar Sands
Tar Sands
 
Posts: 742
Joined: Sat 11 Mar 2006, 04:00:00
Top

Re: Statistics Question

Unread postby jbeckton » Sun 11 Jan 2009, 12:45:28

$this->bbcode_second_pass_quote('emailking', '')$this->bbcode_second_pass_quote('jbeckton', '
')Therefore if the events are NOT independent (such as breaking you leg AND falling down the stairs) then you can NOT reduce P(A and B) to P(A)*P(B).


That's not needed to prove his point though.


Perhaps you are refering to a point of contention between he and Carlhole. Forgive me but I tend to skim through Carlhole's posts at this point....

My contention is and always has been:

P(A)=WTC 7 is on fire
P(B)=WTC collapses

P(A and B) is not P(A) * P(B), rather P(A)*P(B|A)
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00
Top

Re: Statistics Question

Unread postby emailking » Sun 11 Jan 2009, 13:38:38

$this->bbcode_second_pass_quote('jbeckton', '')$this->bbcode_second_pass_quote('emailking', '')$this->bbcode_second_pass_quote('jbeckton', '
')Therefore if the events are NOT independent (such as breaking you leg AND falling down the stairs) then you can NOT reduce P(A and B) to P(A)*P(B).


That's not needed to prove his point though.


Perhaps you are refering to a point of contention between he and Carlhole. Forgive me but I tend to skim through Carlhole's posts at this point....

My contention is and always has been:

P(A)=WTC 7 is on fire
P(B)=WTC collapses

P(A and B) is not P(A) * P(B), rather P(A)*P(B|A)


I would agree with that. Since an collapse inducing explosion is likely to set fire, these are dependent events.
User avatar
emailking
Tar Sands
Tar Sands
 
Posts: 742
Joined: Sat 11 Mar 2006, 04:00:00
Top

Re: Statistics Question

Unread postby Koyaanisqatsi » Mon 12 Jan 2009, 03:40:37

$this->bbcode_second_pass_quote('jbeckton', '')$this->bbcode_second_pass_quote('Koyaanisqatsi', '
')And as usual JBeckton ignores the substance of the post. Dodge and weave!


:lol:

As usual, you presume there is substance in your straw grasping.


Why don't you just answer the question?

$this->bbcode_second_pass_quote('Koyaanisqatsi', 'A')s I said, even if you say you meant both=P(B|A), then the statement doesn't map onto my argument about the 9/11 probabilities, so your claim that this invalidates the argument is itself invalid.
User avatar
Koyaanisqatsi
Lignite
Lignite
 
Posts: 351
Joined: Mon 17 Sep 2007, 03:00:00
Location: Pacific NW
Top

Re: Statistics Question

Unread postby Koyaanisqatsi » Mon 12 Jan 2009, 04:51:45

$this->bbcode_second_pass_quote('emailking', '')$this->bbcode_second_pass_quote('jbeckton', '
')My contention is and always has been:

P(A)=WTC 7 is on fire
P(B)=WTC collapses

P(A and B) is not P(A) * P(B), rather P(A)*P(B|A)


I would agree with that. Since an collapse inducing explosion is likely to set fire, these are dependent events.


This is a strawman. My point has been that the terrorist attack on 9/11 is a unique event with a low probability; and further that WTC 7 fell because of "fire-induced progressive collapse" which, according to NIST, is an entirely new phenomenon that they discovered, which given it's uniqueness also has a very low probability. Thus the probability of them happening together has an even lower probability. Yet JB says that this isn't true:

$this->bbcode_second_pass_quote('jbeckton', '
')$this->bbcode_second_pass_quote('', 'Y')ou are still left with two very unlikely events occurring together, yielding an even smaller probability. As an engineer I think you would know that.

:lol: You are incorrect; the probability of a collapse increases exponentially after the fire has started because the events are not independent. Do you really not see this?

link

It doesn't matter whether one uses the equation for dependent or independent events if P(B) or P(B|A) is sufficiently small:

P(A and B) = P(A)*P(B)
P(A and B) = P(A)*P(B|A)

A = 9/11 terrorist event (causing office fires in adjacent buildings)
B = fire induced progressive collapse of office building

P(A) << 1 (unique event)
P(B) << 1 (unique event)
P(B|A) << 1 (unique event)

It doesn't matter how the problem is framed, multiplying two small numbers together yields a much smaller number. I understand very well that the events are dependent--but this is not a valid counterargument against the co-occurrence of fire-induced progressive collapse of WTC7 and the 9/11 terrorist incident being hugely improbable.

This discussion about probabilities began roughly about here:
link.

By the way, is the statement in the opening post true or false?
User avatar
Koyaanisqatsi
Lignite
Lignite
 
Posts: 351
Joined: Mon 17 Sep 2007, 03:00:00
Location: Pacific NW
Top

Re: Statistics Question

Unread postby jbeckton » Mon 12 Jan 2009, 18:06:43

$this->bbcode_second_pass_quote('Koyaanisqatsi', '
'):lol: You are incorrect; the probability of a collapse increases exponentially after the fire has started because the events are not independent. Do you really not see this?


Yes, the probability of collapse increases after the building is set on fire, therefore:

P(B) does not equal P(B|A)

$this->bbcode_second_pass_quote('Koyaanisqatsi', 'I')t doesn't matter whether one uses the equation for dependent or independent events if P(B) or P(B|A) is sufficiently small:


This is an assumption you have made (you know what happens when you make assumptions). I have done probability (PRA) studies for nuclear power plants that have probabilities 10^(-6) and smaller which is certainly <<1, are you saying we should ignore them? The NRC says they are significant, I agree.

Have you made any attempt to quantify P(B) and P(B|A) before assuming the difference is negligible? I bet you haven’t.

Guess you are just going with “since the probability <<1 it must be impossible".

Well sorry to say that events with a probability <<1 occur everyday. For instance you chances of dying in a plane crash are 11 million to 1 or 9.1e-8

Guess we can assume that no one ever dies in a plane crash since P<<1 right? I wonder what the probability of dying in a office building from an airplane impact would be?

Guess its impossible?
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00
Top

Re: Statistics Question

Unread postby emailking » Mon 12 Jan 2009, 20:04:49

$this->bbcode_second_pass_quote('jbeckton', '
')
Yes, the probability of collapse increases after the building is set on fire, therefore:

P(B) does not equal P(B|A)


He's trying to argue that the probability of the concurrence of two events is smaller than the probability of either event. This is simply true. (unless one of the probabilities is unity in which case equality may hold) Showing this does not depend on assuming P(B) = P(B|A)

It's not hard to see this. In general
P(A and B) = P(A)*P(B|A) <= P(A) since P(B|A) <= 1
But P(A and B) = P(B and A) = P(B)*P(A|B) <= P(B) since P(A|B) <= 1


$this->bbcode_second_pass_quote('jbeckton', '
')This is an assumption you have made (you know what happens when you make assumptions). I have done probability (PRA) studies for nuclear power plants that have probabilities 10^(-6) and smaller which is certainly <<1, are you saying we should ignore them? The NRC says they are significant, I agree.


He's not saying to ignore them (that would entail setting them to 0). If they are less than 1, that constitutes "sufficiently small" for this analysis.

$this->bbcode_second_pass_quote('jbeckton', '
')Have you made any attempt to quantify P(B) and P(B|A) before assuming the difference is negligible? I bet you haven’t.


It doesn't matter what the numbers are as long as all are less than 1.
User avatar
emailking
Tar Sands
Tar Sands
 
Posts: 742
Joined: Sat 11 Mar 2006, 04:00:00
Top

Re: Statistics Question

Unread postby Carlhole » Mon 12 Jan 2009, 20:27:35

$this->bbcode_second_pass_quote('jbeckton', 'P')erhaps you are refering to a point of contention between he and Carlhole. Forgive me but I tend to skim through Carlhole's posts at this point....


What a kunt.
Carlhole
 
Top

Re: Statistics Question

Unread postby jbeckton » Mon 12 Jan 2009, 21:12:55

$this->bbcode_second_pass_quote('emailking', '')$this->bbcode_second_pass_quote('jbeckton', '
')
Yes, the probability of collapse increases after the building is set on fire, therefore:

P(B) does not equal P(B|A)


He's trying to argue that the probability of the concurrence of two events is smaller than the probability of either event.


This is obvious, and not disputed that I know of. The point of contention is to what magnitude is the probability of both smaller than either? Stating that because the probability of both is <<1 means the distinction is negligible is a poor judgment in my estimation. At what point does it become negligible? 1:100, 1:1,000, 1:10000, 1:1000000?

As I just said, the probability of a plane crashing is 1:11000000 but it occurs several times a year so obviously its not impossible, and that is what his analysis would have you believe.

I say there is a big difference between 1:1000 and 1:100000 and treating them as equal because they are <<1 is a mistake, a mistake that will leave your final probability off by several orders of magnitude.

$this->bbcode_second_pass_quote('jbeckton', '
')Have you made any attempt to quantify P(B) and P(B|A) before assuming the difference is negligible? I bet you haven’t.


$this->bbcode_second_pass_quote('emailking', 'I')t doesn't matter what the numbers are as long as all are less than 1.


No?

Say for example with all numbers less than 1:

P(A)=0.5
P(B)=0.005
P(B|A)=0.05

Treating them as independent:

P(A)*P(B)=0.0025 (or 1:400)

Treating them as dependent:

P(A)*P(B|A)=0.025 (or 1:40)

The answer is off by an entire order of magnitude because we used the wrong probability.
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00
Top

Re: Statistics Question

Unread postby jbeckton » Mon 12 Jan 2009, 21:14:10

$this->bbcode_second_pass_quote('Carlhole', '')$this->bbcode_second_pass_quote('jbeckton', 'P')erhaps you are refering to a point of contention between he and Carlhole. Forgive me but I tend to skim through Carlhole's posts at this point....


What a kunt.


Need I say more......
Those that cannot do..... teach. Those that cannot teach......teach gym.-Jack black
User avatar
jbeckton
Expert
Expert
 
Posts: 2082
Joined: Fri 05 Jan 2007, 04:00:00
Top

Re: Statistics Question

Unread postby Carlhole » Mon 12 Jan 2009, 21:35:47

$this->bbcode_second_pass_quote('jbeckton', '')$this->bbcode_second_pass_quote('Carlhole', '')$this->bbcode_second_pass_quote('jbeckton', 'P')erhaps you are refering to a point of contention between he and Carlhole. Forgive me but I tend to skim through Carlhole's posts at this point....


What a kunt.


Need I say more?......


Yeah, say some more kunty stuff.
Carlhole
 
Top

PreviousNext

Return to Open Topic Discussion

Who is online

Users browsing this forum: No registered users and 2 guests

cron